MCPcopy Create free account

hub / github.com/Luce-Org/lucebox-hub / functions

Functions2,878 in github.com/Luce-Org/lucebox-hub

⨍ Functions 2,878◇ Types & classes 420
↓ 1,556 callersMethodsize
server/src/common/gguf_mmap.h:201
↓ 1,435 callersMethoddata
server/src/common/gguf_mmap.h:200
↓ 629 callersMethodempty
server/src/common/moe_hybrid_storage.cpp:199
↓ 285 callersMethodsize
server/src/server/adaptive_keep_ratio.h:93
↓ 217 callersMethodclear
server/src/laguna/laguna_target_graph.cpp:1290
↓ 201 callersFunctionset_last_error
server/src/errors.cpp:17
↓ 192 callersMethodempty
server/src/server/tool_hint.h:40
↓ 164 callersMethodcount
exclusive
server/src/common/score_range.h:12
↓ 152 callersMethod_check
(self, name: str, ok: bool, detail: str = "")
server/tests/test_server_comprehensive.py:52
↓ 145 callersFunctionstep_graph_destroy
Full cleanup: release the persistent gallocr + its CUDA buffer.
server/src/common/step_graph.h:91
↓ 121 callersFunctiondflash27b_last_error
server/src/errors.cpp:24
↓ 114 callersMethodemit
server/src/common/daemon_loop.cpp:35
↓ 103 callersMethodencode
server/src/server/tokenizer.cpp:636
↓ 87 callersMethodemit_token
server/src/server/sse_emitter.cpp:225
↓ 81 callersFunctionactivation_pair_free
server/src/common/dflash_layer_split_runtime.h:76
↓ 81 callersFunctionread_exact_fd
server/src/common/io_utils.h:98
↓ 75 callersMethodwrite
server/test/test_drafter_warm_path_regression.cpp:83
↓ 74 callersMethoddecode
server/src/server/tokenizer.cpp:780
↓ 66 callersFunctionstream_status
server/src/common/dflash_draft_ipc.h:89
↓ 63 callersFunctionwrite_exact_fd
server/src/common/io_utils.h:113
↓ 59 callersMethodembed
server/src/qwen35/gguf_target_loader.cpp:82
↓ 52 callersFunctionsample_logits
server/src/common/sampler.cpp:107
↓ 47 callersFunctionds4_elapsed_us
server/src/deepseek4/deepseek4_graph.cpp:39
↓ 43 callersMethodreset
server/src/common/moe_expert_compute.cpp:153
↓ 40 callersMethod_report
(self, name: str, ok: bool, detail: str = "")
server/tests/test_server_smoke.py:41
↓ 39 callersMethodemit_finish
server/src/server/sse_emitter.cpp:498
↓ 37 callersFunctionparse_tool_calls
server/src/server/tool_parser.cpp:574
↓ 37 callersFunctionstream_emit
server/test/pflash_daemon.cpp:51
↓ 36 callersFunctionnow_ms
server/test/test_kvflash.cpp:49
↓ 35 callersFunctionpost_json
(path, body, timeout=GEN_TIMEOUT)
server/scripts/test_server_integration.py:41
↓ 34 callersMethodactive
server/src/qwen35/qwen35_target_shard_ipc.h:84
↓ 32 callersFunctionactivation_buffer_free
server/src/common/dflash_layer_split_runtime.h:86
↓ 31 callersMethod_req
(self, method: str, path: str, body: dict | None = None, stream: bool = False, timeout: float = 1
server/tests/test_server_comprehensive.py:41
↓ 31 callersMethodenabled
server/src/placement/remote_draft_config.h:14
↓ 31 callersFunctionfree_qwen35_layer_split_shards
server/src/qwen35/layer_split_forward.cpp:665
↓ 31 callersFunctionfree_target_cache
server/src/qwen35/qwen35_target_graph.cpp:286
↓ 31 callersMethodupdate
server/src/server/adaptive_keep_ratio.h:56
↓ 29 callersMethodopen
server/src/common/gguf_mmap.h:129
↓ 28 callersFunctionapply_scale2
NVFP4 scale2: if weight has a per-tensor scale, multiply the matmul result by that scale. No-op when scale==1.0f (non-NVFP4 models).
server/src/common/moe_hybrid_ffn_eval.cpp:19
↓ 28 callersMethodemit_start
server/src/server/sse_emitter.cpp:151
↓ 28 callersFunctionfree_draft_weights
server/src/draft/draft_safetensors_loader.cpp:694
↓ 28 callersFunctionfree_drafter
server/src/qwen3/qwen3_drafter.cpp:245
↓ 27 callersFunction_w
Get weight tensor, ensuring it's in the target half dtype.
optimizations/megakernel/model.py:86
↓ 26 callersFunctioncheck
server/test/test_restore_delta.cpp:9
↓ 26 callersMethodneeds_logit_processing
True when any logit modifier is active (penalties or stochastic sampling). Backends should use CPU sample_logits() path when this returns true.
server/src/common/sampler.h:38
↓ 26 callersFunctionpipe_elapsed_us
server/src/qwen35moe/qwen35moe_pipelined_decode.cpp:23
↓ 26 callersMethodslot_for
Physical pool slot for logical position `pos`. Allocates (and, when the pool is full, evicts) at chunk granularity. Call once per appended token, in l
server/src/common/kvflash_pager.h:234
↓ 25 callersFunctionalign_up
server/src/common/attn_masks.h:22
↓ 25 callersFunctionconcat
Concatenate all SSE chunks into a single string.
server/test/test_server_unit.cpp:163
↓ 25 callersFunctionmake_emitter
server/test/test_server_unit.cpp:156
↓ 25 callersFunctionms
server/src/deepseek4/deepseek4_backend.cpp:76
↓ 24 callersFunctiong4_free_pers
server/src/gemma4/gemma4_graph.cpp:1194
↓ 24 callersFunctionread_u32
server/src/server/disk_prefix_cache.cpp:182
↓ 24 callersFunctionrm_rf
Helper: recursively remove a directory.
server/test/test_server_unit.cpp:2417
↓ 23 callersFunctiondraft_feature_mirror_free
server/src/common/dflash_feature_ring.cpp:233
↓ 23 callersMethodfree
server/src/deepseek4/deepseek4_graph.cpp:241
↓ 23 callersFunctionget_u32_or
server/src/deepseek4/deepseek4_loader.cpp:72
↓ 23 callersFunctionpick
(*keys)
server/scripts/convert_dflash_to_gguf.py:82
↓ 22 callersFunctionfree_target_weights
server/src/qwen35/gguf_target_loader.cpp:822
↓ 22 callersFunctiongemma4_rms_norm_mul
server/src/gemma4/gemma4_graph.cpp:39
↓ 22 callersFunctiongguf_tensor_in_file
True iff the tensor's data region lies fully within the mapped file. data_off: byte offset of the GGUF data section (gguf_get_data_offset) tensor_of
server/src/common/gguf_bounds.h:32
↓ 22 callersFunctionload_drafter
server/src/qwen3/qwen3_drafter.cpp:132
↓ 22 callersFunctionmoe_expert_compute_elapsed_us
server/src/common/moe_expert_compute_ipc.cpp:52
↓ 22 callersFunctionstep_graph_free
Reset the per-call graph state (ctx + graph + tensor handles) but KEEP the persistent CUDA buffer in `sg.alloc` alive across steps.
server/src/common/step_graph.h:67
↓ 21 callersFunctionapply_scale2
NVFP4 scale2: if weight has a per-tensor scale, multiply the matmul result by that scale. No-op when scale==1.0f (non-NVFP4 models).
server/src/qwen35/qwen35_ops.h:20
↓ 21 callersMethodchunk_tokens
server/src/common/kvflash_pager.h:205
↓ 21 callersFunctionelapsed_us
server/src/common/moe_hybrid_ffn_eval.cpp:38
↓ 21 callersFunctionfree_laguna_target_weights
server/src/laguna/laguna_target_loader.cpp:689
↓ 21 callersFunctionget_u32_or
server/src/qwen35/gguf_target_loader.cpp:117
↓ 21 callersFunctionget_u32_or
server/src/laguna/laguna_target_loader.cpp:74
↓ 21 callersFunctionsocket_close
server/src/server/http_server.cpp:57
↓ 21 callersFunctiontensor_ready
server/src/laguna/laguna_layer_split_adapter.cpp:33
↓ 21 callersFunctiontensor_ready
server/src/gemma4/gemma4_layer_split_adapter.cpp:27
↓ 20 callersMethodalloc_span
Allocate slots for [kv_start, kv_start + n_tok) ahead of a forward step (evicting LRU/low-score chunks as needed). False — with a diagnostic — if the
server/src/common/kvflash_pager.h:218
↓ 20 callersMethodattach
server/src/common/kvflash_pager.h:101
↓ 20 callersFunctionexpect
server/test/test_qwen35moe_routing_stats.cpp:11
↓ 20 callersMethodfree
server/src/qwen35moe/qwen35moe_pipelined_decode.cpp:29
↓ 20 callersMethodis_layer_split
server/src/placement/placement_config.h:30
↓ 20 callersFunctionnew_like_with_expert_count
server/src/common/moe_hybrid_storage.cpp:133
↓ 20 callersFunctionparse_kv_type
server/src/kv_quant.cpp:34
↓ 20 callersFunctionraxGetData
Get the node auxiliary data. */
server/src/server/rax.c:284
↓ 19 callersMethodcommand_stream
server/src/common/backend_ipc.h:103
↓ 19 callersMethodrun
server/src/server/http_server.cpp:1090
↓ 19 callersMethodstream_fd
server/src/common/backend_ipc.h:105
↓ 18 callersMethod_report
(self, name: str, ok: bool, detail: str = "")
server/tests/test_tokenizer.py:94
↓ 18 callersMethoddestroy
server/test/test_kvflash.cpp:205
↓ 18 callersFunctionenv_int
server/src/qwen3/qwen3_drafter.cpp:61
↓ 18 callersFunctionexpect
server/test/test_kvflash_placement.cpp:15
↓ 18 callersMethodget_keep_ratio
server/src/server/adaptive_keep_ratio.h:69
↓ 18 callersFunctionget_u32_or
server/src/gemma4/gemma4_loader.cpp:95
↓ 18 callersFunctionhash_prefix
server/src/server/prefix_cache.cpp:128
↓ 18 callersMethodkill
(self)
optimizations/spark/spark/_daemon.py:95
↓ 18 callersFunctionlayer_split_shard_metas
server/src/common/layer_split_utils.h:47
↓ 18 callersFunctionnearly_equal
server/tests/test_deepseek4_unit.cpp:52
↓ 18 callersFunctionparse_nonnegative_int
server/src/ipc/backend_ipc_main.cpp:48
↓ 18 callersMethodreset
(self)
optimizations/megakernel/model.py:297
↓ 18 callersMethodstep
server/test/test_kvflash.cpp:184
↓ 17 callersFunctionactivation_pair_init
server/src/common/dflash_layer_split_runtime.h:95
↓ 17 callersFunctioncreate_target_cache
server/src/qwen35/qwen35_target_graph.cpp:74
↓ 17 callersMethoddecode
(self, ids: list[int])
server/tests/test_tokenizer.py:62
next →1–100 of 2,878, ranked by callers